Hi everyone,

I’m seeing an intermittent issue with a ChatGPT App Store app that exposes multiple MCP tools.

The expected flow is that ChatGPT should call several tools sequentially in the same conversation. For example, the first tool call returns some basic instructions, and the following tool calls continue the workflow.

What I’m seeing instead:

1. The first tool call is available and gets called successfully.
2. After that, ChatGPT is unable to call the next tools.
3. When I ask ChatGPT to call the next tool, it says those tools are not available / not exposed as callable tools in this chat.
4. This happens for roughly half of user sessions.
5. It eventually stops happening on its own, but refreshing the app or reconnecting does not seem to fix it.

This looks similar to the issue described here, however our issue started at around 21:00 UTC on the 25th of June:

Summary A custom Remote MCP server (connected as a ChatGPT app via Developer Mode; OAuth 2.1 + DCR + PKCE) works fine in fresh conversations, but in long-lived or older conversations — including threads inside a Project — its tools stop being exposed to the model for that conversation. The model can name the tools but can’t call them, and no request reaches our server. A new chat restores everything. Is this expected, and is there a way to detect or recover from it? Symptom (reproducible, same …

A few extra notes:

• The tools are definitely defined and exposed by the MCP server.
• When I open the manifest in the settings tab, I see all the tools.
• No server side issues, the first tool call is returned successfully.
• The model can sometimes describe or remember the tool names, but says they are not callable in the current chat.
• Waiting eventually resolves it, but reconnecting/refreshing does not help.

Has anyone else seen this with ChatGPT apps? Any guidance on whether this is a platform-side issue would be appreciated.

Thanks!

I wonder if this is related to the new version of GPT-5.5 Instant released last week. Can anyone from OpenAI confirm whether Apps on Instant have a smaller effective context or tool-descriptor budget? I saw docs implying context size for Instant is now 16K tokens (and it used to be 27K tokens).

Specifically, can large MCP tools/list payloads - descriptions, input/output schemas, annotations, metadata, etc. - cause exposed tools to become unavailable or stop being selected after an initial tool call?